Table 1 Performance metrics (ROC–AUC and AUC–PR) for personalized federated learning models compared to federated learning baseline across various strategies

From: Personalized federated learning for predicting disability progression in multiple sclerosis using real-world routine clinical data

 

ROC–AUC ↑

AUC–PR ↑

 

Personalized FL

FL

Personalized FL

FL

Experiments

Fine-tuned

Adaptive

Baseline

Fine-tuned

Adaptive

Baseline

FedAVG

0.8370 ± 0.0016

0.8384 ± 0.0014

0.7840 ± 0.0019

0.5156 ± 0.0046

0.5290 ± 0.0062

0.4030 ± 0.0059

FedProx

0.8375 ± 0.0019

0.8398 ± 0.0019

0.7834 ± 0.0019

0.5221 ± 0.0044

0.5346 ± 0.0029

0.4081 ± 0.0058

FedAdagrad

0.8340 ± 0.0012

0.8361 ± 0.0021

0.7762 ± 0.0021

0.5043 ± 0.0043

0.5131 ± 0.0062

0.3913 ± 0.0061

FedYogi

0.8369 ± 0.0027

0.8178 ± 0.0026

0.7910 ± 0.0028

0.5379 ± 0.0072

0.4702 ± 0.0059

0.4420 ± 0.0078

FedAdam

0.8339 ± 0.0015

0.8324 ± 0.0032

0.7920 ± 0.0031

0.5383 ± 0.0050

0.5197 ± 0.0073

0.4488 ± 0.0061

Centralized

 

0.8092 ± 0.0012

  

0.4605 ± 0.0043

 
  1. “Centralized” results are included for comparison purposes and do not fall under the FL or PFL categories. For brevity, the term “AdaptiveDualBranchNet” will be referred to simply as “Adaptive” throughout this manuscript. Additionally, the non-personalized FL model is commonly referred to as the baseline FL paradigm. The value after “±” denotes the standard deviation of the measurements. The bold values indicate the best-performing results within each row, where higher values are better, as shown by the arrows in the column headers.
Back to article page